HAL Id: hal-00769625
https://hal.inria.fr/hal-00769625
Submitted on 2 Jan 2013
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Effectively long-distance dependencies in French :
annotation and parsing evaluation
Marie Candito, Djamé Seddah
To cite this version:
Marie Candito, Djamé Seddah. Effectively long-distance dependencies in French : annotation and parsing evaluation. TLT 11 - The 11th International Workshop on Treebanks and Linguistic Theories, Nov 2012, Lisbon, Portugal. �hal-00769625�
Effectively long-distance dependencies in French :
annotation and parsing evaluation
Marie Candito⋆and Djamé Seddah⋆!
⋆Alpage (Univ. Paris Diderot & INRIA), 175 rue du Chevaleret, 75013 Paris, France !
Univ. Paris Sorbonne, 28, rue Serpente, 75006 Paris, France
marie.candito@linguist.jussieu.fr, djame.seddah@paris-sorbonne.fr
Abstract
We describe the annotation of cases of extraction in French, whose previous annotations in the available French treebanks were insufficient to recover the correct predicate-argument dependency between the extracted element and its head. These cases are special cases of LDDs, that we call effectively long-distance dependencies (eLDDs), in which the extracted element is indeed separated from its head by one or more intervening heads (instead of zero, one or more for the general case). We found that extraction of a dependent of a finite verb is very rarely an eLDD (one case out of 420 000 tokens), but eLDDs corresponding to extraction out of infinitival phrase is more fre-quent (one third of all occurrences of accusative relative pronoun que), and eLDDs with extraction out of NPs are quite common (2/3 of the occurrences of relative pronoun dont). We also use the annotated data in statistical depen-dency parsing experiments, and compare several parsing architectures able to recover non-local governors for extracted elements.
1
Introduction
While statistical parsers obtain high overall performance, they exhibit very differ-ent performance across linguistic phenomena. In particular, most statistical parsers perform poorly on long-distance dependencies (LDDs), which, though rare, are important to fully recover predicate-argument structures, which are in turn needed for semantic applications of parsing. Poor performance on LDDs is known of English statistical parsers, even though the training data does contain information for resolving unbounded dependencies (the Penn Treebank, or the specific dataset evaluated by Rimell et al. [17]). For French, the situation is worse, since the usual training data, the French Treebank (Abeillé and Barrier [1]), is a surface syntag-matic treebank that does not contain indications of LDDs : extracted elements bear a grammatical function, but no annotation indicates their embedded head. Hence syntagmatic stochastic French parsers cannot capture LDDs. Concerning depen-dency parsing, French dependepen-dency parsers can be learnt on the DEPFTB, resulting
from the automatic conversion of the FTBinto projective dependency trees (Can-dito et al. [3]). But we will show that this automatic conversion leads to wrong dependencies for particular cases of LDDs - cases we call effectively-long-distance dependencies (eLDDs), in which the fronted element is extracted from an actually embedded phrase -, and thus statistical parsers learnt on the DEPFTBare unable to recover such cases correctly.
In this paper, we describe the manual annotation, performed on the FTBand
the Sequoia treebank (Candito and Seddah [5]), of the correct dependencies in eLDDs, leading to non-projective dependency treebanks. We then evaluate several dependency parsing architectures able to recover eLDDs.
2
Target linguistic phenomena
Extraction is a syntactic phenomena, broadly attested across languages, in which a word or phrase (the extracted element) appears in a non-canonical position with respect to its head. For a given language, there are well-defined contexts that in-volve and licence an extraction. In English or French, the non-canonical position corresponds to a fronting of the extracted element.
A first type of extraction concerns the fronting of the dependent of a verb, as in the four major types topicalization (1), relativization (2), questioning (3), it-clefts (4), in which the fronted element (in italics) depends on a verb (in bold):
(1) À nos arguments, To our arguments, (nous (we savons know que) that) Paul Paull opposera will-oppose les his. siens. (2) Je I connais know l’ the homme man que that (Jules (Jules pense thinks que) that) Lou Lou épousera. will-marry. (3) Sais-tu Do-you-know à qui to whom (Jules (Jules pense thinks que) that) Lou Lou donnera will-give un a cadeau? present? (4) C’ It est is Luca Luca que that (Jules (Jules pense thinks que) that) Lou Lou épousera. will-marry.
Since transformational grammar times, any major linguistic theory has its own ac-count of these phenomena, with a vocabulary generally bound to the theory. In the following we will use ’extracted element’ for the word or phrase that appears fronted, in non canonical position. One particularity of extraction is ’unbounded-ness’: stated in phrase-structure terms, within the clause containing the extraction, there is no limit to the depth of the phrase the extracted element comes from. In dependency syntax terms, for a fronted element f appears either directly to the left of the domain of its governor g, or to the left of the domain of an ancestor of g. We focus in this paper on the latter case, that we call effectively long-distance de-pendencies (eLDD). In examples (1) to (4), only the versions with the material in brackets are eLDDs.
In French, another case of eLDD is when a PP is extracted from a predicative complement either nominal or adjectival:
(5) la the mélancolie melancholy à to laquelle which il he est is enclin
prone (’the melancholy he is prone to’) Other cases of eLDDs arise for some PPs with preposition de, which under some well-studied conditions (see for instance Godard [8]) can be extracted from subject or direct object NPs, as in (6). (6) un a échec failure dont of-whom (Léo (Léo me to-me dit says que) that) les the causes causes sont are bien well connues known ’a failure whose causes (Leo tells me) are well known’
Those de-phrases are precisely the ones that can be cliticized, with the anaphoric clitic en (of-it) appearing on the verb governing the NP, as in (7).1
(7) Tu you en of-it connais know bien well les the raisons
reasons ’You know well the reasons for it’ To sum up, eLDDs comprise any case of extraction out of predicative comple-ments, nominal subjects or objects (examples (5) to (7)), and cases of extraction of a dependent of a verb (examples (1) to (4)) only when involving intervening heads, (such as to think in these examples).
3
Target French Treebanks
3.1 French treebank and Sequoia treebank
Our objective is to obtain a dependency treebank for French with correct governors for extracted elements. This treebank will thus contain non-projective links. We perform our annotation of cases of extraction on two treebanks :
• the French Treebank (Abeillé and Barrier [1]) (hereafter FTB), a constituency
treebank made of 12351 sentences2from the national newspaper Le Monde • the Sequoia treebank (Candito and Seddah [5]), an out-of-domain corpus
annotated following the FTB’s annotation scheme. It contains roughly 1000
sentences from the French wikipedia, 1000 sentences from the medical do-main (from medicines’ marketing authorization reports from the European Medecine Agency), 500 sentences from Europarl and 500 sentences from the regional newspaper L’Est Républicain.
These are constituency treebanks, in which the dependents of verbs are labeled with a grammatical function, since a given structural position may correspond to different grammatical relations. Candito et al. [3] describe a tool for the automatic
1Note that in that case, the dependency between the clitic and the noun is bounded, but we still
consider it a eLDD, because the clitic appears locally to the head (the verb) of the NP it depends on.
conversion of such constituency trees into surface dependency trees.3 We briefly describe below this procedure, and detail the incorrect result obtained for eLDDs.
3.2 Automatic conversion to surface dependencies
The conversion procedure is based on the classic technique of head propagation rules, proposed for English by Magerman [10], and outputs projective surface de-pendency trees (each token has exactly one governor, except the root) : (i) Nodes in phrase-structure trees are annotated with their lexical head, using head-propagation rules, that state how to find the syntactic head in the right-hand side of a CFG rule; (ii) Using the lexical heads, bilexical dependencies are extracted. If the constituent node for the dependent bears a functional label, it is used as the label of the depen-dency; (iii) Remaining unlabeled dependencies are labeled using heuristics.
With that technique, output dependency trees are necessarily projective, and non-local dependents are wrongly attached to the local lexical head, as exemplified in Figure 1 in which the accusative relative pronoun que is wrongly attached to the local head semblaient (seemed) instead of its actual syntactic head partager (to
share). A crucial point here is that a wrong dependency arises only for LDDs that
are eLDDs. For instance from a simplified version of tree (a), without the raising verb, we would obtain the correct dependency between que and the verb partager.4
(a) NP D un N sentiment Srel NP-OBJ PROREL que NP PRO tous VN V semblaient VPinf-ATS VN VINF partager (b)
un sentiment que tous semblaient partager DET MOD_REL
OBJ SU
J ATS
Figure 1: Left: An NP as annotated in the original French Treebank scheme (for
a feeling that (they) all seemed (to) share). Right: corresponding automatically
derived dependency tree, with wrong governor for the wh-word que.
4
Manual annotation
4.1 Selection of occurrences to annotate
One major difficulty to annotate extractions is that though ubiquitous in the linguis-tic literature, they are quite rare in actual texts. Further, some cases like topical-ization (cf. example (1) above), involve structural clues only, and no lexical clue, namely no wh-words. Topicalization does exist in French, but is much rarer than
3Included in the BONSAItoolkit (http://alpage.inria.fr/statgram/frdep/fr_stat_dep_parsing.html). 4Note that the anaphoric relation between the pronoun and its antecedent is then trivial to recover,
relativization, it-clefts or questioning, and is restricted to the extraction of prepo-sitional phrases. Hence, because we could not afford to scan the whole treebank to look for extractions, and also because it is unclear whether such dependencies can be recovered with current parsing techniques, we chose as a first step to fo-cus on words known to be likely to involve an extraction, namely the clitic en and wh-words (relative and interrogative pronouns and determiners). This allows to capture the cases exemplified in section 2, except topicalization.
We examined each occurrence of wh-word and of the clitic en, and annotated the correct dependency in case of non-local dependency.
4.2 Annotation scheme and methodology
4.2.1 Functional paths
We chose to annotate non-locality using functional paths, made of sequences of de-pendency labels, as proposed in the LFG framework. We were inspired by Schluter and van Genabith [18], who have used them to annotate a manually modified ver-sion of half the French Treebank.5Before formalizing this notion, let us take as ex-ample the automatically-derived dependency tree in Figure 1, in which the relative pronoun que is wrongly attached to the local governor semblaient. The non-local governor is partager. We define the functional path for que to be the sequence of la-bels appearing on the path between que and its correct governor, namely OBJ.AT S, which can be read as “’que’ should be the OBJ of the ATS of its local governor”.
More generally, let d be a lexical item that should depend on a non-local gov-ernor nlg. Let ADT be the dependency tree obtained by automatic conversion from the source constituency tree, and CDT the correct dependency tree that we target. Since nlg is non-local, the tree ADT contains a wrong dependency lg→ d, whilel0
CDT contains nlg→ d (we suppose here that the label ll0 0 is correct is ADT ). For
such cases, we define a functional path as the sequence of labels l0.l1....ln that
appear, in the incorrect tree ADT , on the path between the dependent d and its non-local governor nlg.
The manual annotation consists in making functional paths explicit : in the previous formalization, it amounts to modifying ADT into ADT′, by labeling the dependency lg−→ d with the functional path as label : lgl0 l0.l1...ln
−→ d. Note that eLDDs are exactly the cases involving a functional path of length > 1 instead of a simple functional tag (which can be regarded as a functional path of length 1).
4.2.2 Automatic interpretation of functional paths
This kind of annotation can be used in a straightforward way to recover the correct governor of extracted elements. We give in figure 2 the algorithm used to interpret functional paths as the indication of non-local dependencies : it changes a tree
5But these authors obtain only 65 cases in total, suggesting the linguistic constructions covered
containing functional paths into a corresponding tree, in which dependents that are non-local, are attached to their non-local governors.
0 • ADT′← ADT
1 • while ADT′contains a dependency a of the form lgl0−→ d, with n.l1...ln > 0, do 2 • i ← n, gi← lg
3 • while i > 0 do
4 • G ← nodes x such as gi li
−→ x and x &= d
5 • if G &= /0, choose the left-most most appropriate node x, and set gi−1← x 6 • else, continues to next element in while 1
7 • i ← i − 1
8 • replace in ADT′the dependency a by g0
l0 −→ d
Figure 2: Algorithm to interpret functional paths : find the non-local governors and change the dependency tree accordingly
If we go back to the example of Figure 1, the manual annotation applied to tree (b) is given in tree (c) in Figure 3. The interpretation of the sole functional path in tree (c) outputs the tree (d), which is non-projective.
The functional paths can be inconsistent with respect to the tree they appear in. This results in obtaining an empty set at line 4 in Figure 2. During the manual annotation phase, such cases can be used as warnings for a wrong functional path. When using the procedure in the parsing phase (see section 5), such functional paths are simply discarded and the label l0.l1...lnis replaced by l0.
Further, the functional paths can be ambiguous. This is the case when the set
Gobtained at line 4 contains more than one element, at any point in the functional path. In these cases, the procedure prefers verbal governors over any other POS, and leftmost governor in case of remaining ambiguity.
This procedure is similar to the deprojectivization procedure defined by Nivre and Nilsson [15], when these authors use the encoding scheme they called ’Head’. More precisely, both procedures are equivalent in case of a functional path of length 2. We needed to define a procedure able to cope with paths of arbitrary length, in order to interpret any manually annotated functional path.
(c)
un sentiment que tous semblaient partager DET MOD_REL OBJ.AT S SU J ATS (d)
un sentiment que tous semblaient partager DET MOD_REL
OBJ
SUJ ATS
Figure 3: Left: Tree (c) is a modification of tree (b), with (manually) annotated functional path for the extracted element que. Right: Tree (d) is the output (non-projective) tree after using the algorithm Figure 2 to interpret functional paths.
4.2.3 Annotation methodology
Because the original treebanks we use have been annotated in constituency format, we chose to annotate the above-mentioned functional paths in the constituency trees, in order to retain the property that the dependency treebank can be automat-ically converted from the constituency trees. So, for the example of Figure 1, we annotate the functional path OBJ.AT S on the NP node that dominates the relative pronoun : NP-OBJ is replaced by NP-OBJ.ATS, then the usual constituency-to-dependency conversion produces the left tree of Figure 3, and the functional path interpretation procedure produces the right tree of Figure 3.6
We performed manual annotation of functional paths on a bracketed version of the French Treebank (called the FTB-UC by (Candito et al. [3])), and on the bracketed version of the Sequoia Treebank, using the WordFreak tool (Morton and LaCivita [13]), customized to handle the relevant sets of POS, non terminals and functional labels. The annotation for the words en and dont were performed in-dependently by two annotators using WordFreak, and adjudicated by an expert annotator. The annotation for all the other wh-words were directly performed by a single expert annotator, because they potentially involve longer dependencies than for the word en, hence requiring to define more difficult (longer) functional paths. Then we applied the functional path interpretation algorithm (Figure 2), to obtain dependency versions of the FTBand the SEQTBwith correct governors in case of eLDD. The resulting annotations are freely available.7
4.3 Quantitative characteristics
The resulting annotation provides a picture of the prevalence of extraction phe-nomena for a corpus of journalistic text (FTB) and for a corpus with mixed genres (the Sequoia corpus). We give various numbers of tokens for the annotated data in table 1, for the FTB, the SEQTB, and for the concatenation of both corpora.
The first observation is that the cases of extraction leading to eLDDs are very rare : for the whole FTB+ SEQTBcorpus, 0.16% of the tokens received a non-local governor (i.e. a functional path of length> 1). Further, more than 80% of eLDDs have a functional path of length 2, namely are "not-so-long" distance dependencies. Focusing on projectivity, we see that only around two thirds of the eLDDs are non-projective (359 out of 618 dependencies). This is because most eLDDs are extractions from a subject NP, as in (6) (noted with functional path ’DEP.SUJ’ : the extracted element is the dependent of the subject of the local (wrong) head). In
6We are aware that such an annotation task supposes a very good understanding of both the
linguistic phenomena at play, and of the ad-hoc conversion to dependency procedure. Yet this has the advantage, over more traditional annotation using coindexed traces, that annotation on bracketed constituency trees is easier than on dependency structure.
7The S
EQTB with correctly annotated eLDDs is available at https://www.rocq.inria.fr/alpage-wiki/tiki-index.php?page=CorpusSequoia. The FTBwith non-local dependencies is available on re-quest, provided you have the license for the FTB.
general, no dependent of the local verbal head intervenes between the subject and the extracted element (en or dont) as in (6), projectivity is preserved.8
Number of tokens
Treebank total with eLDD (%) fpl=2 fpl=3 fpl>3 non projective
FTB 350931 555 (0.16 %) 466 69 20 317
SEQTB 69238 63 (0.09 %) 47 13 3 42
FTB+ SEQTB 420169 618 (0.15%) 513 82 23 359
Table 1: For the FTBand the SEQTB, total number of tokens, tokens with non-local dependency, i.e. with length of functional path (fpl) > 1, tokens with fpl=2, fpl=3, fpl > 3; Number of tokens with fpl>1 that have a non projective dependency.
If we focus on the top three lexical items that exhibit eLDDs, we obtain the ac-cusative relative pronoun que, the genitive relative pronoun dont and the anaphoric clitic en. As can be seen in table 2, these three lemmas totalize 570 out of the 618 cases of annotated eLDD in the FTB+ SEQTBtreebanks.9 Note that though we saw
eLDDs are very rare, these particular three lemmas have a substantial proportion of eLDD occurrences, especially dont (65.7% of its occurrences).
The most frequent element extracted from a verbal phrase is the relative pro-noun que. A striking observation is that for all the 152 eLDD cases involving que, the embedded phrase que originates from is infinitival (as in the example Figure 3), and none is a finite clause.10
However, though the relative pronoun dont and the clitic en can either depend on verbs, nouns or adjectives, the majority of eLDDs are cases of extraction out of NPs. More precisely, out of subject NPs for dont (251 cases of functional paths DEP.SUJ out of 329) and out of object NPs for en (51 cases of functional paths DEP.OBJ, out of 89). The other prevalent case for clitic en is the extraction out of predicative complement (24 cases of functional paths DEP.ATS).
5
Parsing experiments
In this section we list and evaluate various parsing strategies able to output eLDDs. Though the overall parsing performance is unlikely to be affected by this parameter (given the very small amount of annotated eLDDs), it seems interesting to focus on attachment scores for the specific tokens that do often trigger eLDDs.
We relate experiments both using the “local” dependency trees, namely trees automatically converted from constituency trees without functional paths, and
us-8Non projectivity also arises when extracting a more deeply embedded dependent within the
subject NP, as in “une entreprise dont la moyenne d’âge des salariés dépasse 40 ans” (a company of-which the average of age of the employees is over 40).
9Other cases concern (i) extractions of prepositional phrases (pied-piping), such as in example 3,
for which the token that will bear the eLDD is the head preposition of the PP (à in example 3) or (ii) rare cases of extraction of NPs with wh-determiner, for which the noun bears the eLDD.
10We found no extraction out of an embedded finite clause in the F
TB, and one in the SEQTB, in which the extracted element is a PP.
Number of occurrences in FTB+ SEQTB
Lemma POS total local gov non-local gov Top-3 most frequent fct paths
que relative 616 464 152 (32,8%) OBJ.OBJ 77
pronoun OBJ.OBJ.OBJ 22
OBJ.OBJ.DE_OBJ 19
dont relative 501 172 329 (65.7%) DEP.SUJ 251
pronoun DEP.OBJ 29
DEP.ATS 27
en accusative 411 322 89 (21,7%) DEP.OBJ 51
clitic DEP.ATS 24
DEP.SUJ 11
Table 2: Statistics for the three lexical items having a non-local governor most frequently: total number of occurrences, number with and without local governor, and top-three most frequently-annotated functional paths.
ing the “non local” dependency trees, which have corrected dependencies for eL-DDs (obtained via the procedure described in section 3.1). The local trees are pro-jective, whereas the non local trees contain a few non projective links (two thirds of the eLDDs are non projective, cf. section 4.3).
We use the usual split for the 2007 version of the FTB(1235, 1235 and 9881 sentences for test, development and training sets respectively). For evaluation, we use as test sets the whole SEQTB on top of the usual FTB development and test
sets.11 In all our experiments the predicted parses are evaluated against the non
local(pseudo-)gold dependency trees (while for training, we sometimes use the
localdependency trees). We provide unlabeled and labeled attachment scores for all tokens except punctuation, and for the three lexical items that have a non local governor most frequently (the three lemma+POS pairs of Table 2).
We use MaltParser (Nivre et al. [14]), version 1.7 and MSTParser (McDonald [11]), version 0.5.0.12 For the features, we use the best ones from a benchmarking of dependency parsers for French (Candito et al. [4]), except we removed unsu-pervised word clusters as features.13 For MaltParser, we always use the arc-eager
parsing algorithm, that can provide projective trees only.14 For MSTParser, we ei-ther use order 1 factors and the Chu-Liu-Edmonds algorithm, that can output non projective trees, or order 2 factors, with the Eisner algorithm, restricted to projec-tive trees. We also test pseudo-projecprojec-tive parsing (Nivre and Nilsson [15]), where non projective trees are transformed into projective ones, with label marking to allow the reverse the graph transformation: we use MaltParser to obtain projec-tivized versions of the non local trees, then either MaltParser or MSTParser to train a (projective) parser on these, and MaltParser again to de-projectivize the obtained
11As we did no parameter tuning, we did not reserve a test set for final test.
12Available at http://www.maltparser.org and http://sourceforge.net/projects/mstparser. 13POS are predicted using M
ORFETTE (Chrupa!a et al. [6]), and lemmas and morphological features are predicted using the BONSAItoolkit.
14Experiments with the swap algorithms showed degraded performance, supposedly due to the
predicted parses.15
SEQTB FTBdev FTBtest
LAS UAS UAS on LAS UAS UAS on LAS UAS UAS on
Type of the 216 the 126 the 174
training Parsing que, en que, en que, en
Parser trees algo dont dont dont
MALT local arc-eager 85.0 88.0 57.9 (125) 86.6 89.0 55.2 (75) 87.3 89.6 54.0 (94)
MALT pproj arc-eager 84.9 88.0 77.3 (167) 86.7 89.1 77.2 (105) 87.4 89.7 76.4 (133)
MST non local o1 np 85.0 88.2 71.8 (155) 86.5 89.2 80.2 (109) 87.3 89.9 80.5 (140)
MST local o2 proj 85.6 88.9 56.0 (121) 87.6 90.3 58.8 (80) 88.2 90.8 56.3 (98)
MST non local o2 proj 85.6 89.0 68.1 (147) 87.5 90.2 74.3 (101) 88.1 90.7 73.0 (127)
MST pproj o2 proj 85.7 89.1 71.3 (154) 87.6 90.3 80.9 (110) 88.4 91.0 78.7 (137)
Table 3: Parsing performance for malt or mst, with training on either local, non local, or projectivized (pproj) trees.
If we look at the overall performance, we observe, as usual, that MSTParser order 2 performs better than MaltParser. Further, training on the local, non local or projectivized trees has practically no impact (differences are not significant). When focusing on the lexical items that are likely to exhibit a non local dependency, the picture is quite different. First, it can be noted that using the non projective decod-ing for MSTParser, because it implies usdecod-ing order 1 factors, is not worth: the gain obtained for the eLDDs is too small to counterbalance the drop in performance when using order 1 instead of order 2 factors. Second, when comparing perfor-mance across the various training trees types, we note that pseudo-projectivization performs best on eLDDs. Moreover performance on eLDDs with training on pro-jectivized data is comparable for MaltParser and MSTParser. The best strategy both overall and for eLDDs seems to be to use pseudo-projectivization with MSTParser and projective decoding, in order to use higher order factors.
6
Related Work
The Penn Treebank’s annotation style of all types of long distance dependencies considerably eases the extraction of wide coverage grammars, and the building of high performing deep syntax parsers, as long as their underlying linguistic frame-work is able to cope with those complex phenomena (Hockenmaier et al. [9] for CCG; Cahill et al, [2] for LFG based parsing; Miyao et al. [12] for HPSG). Depen-dency parsers, when trained on a non projective depenDepen-dency version of the PTB,
and with the use of post-parsing heuristics, exhibit a similar range of performance than the parsers cited above (Nivre et al. [16]). For French, as noted in introduction, non local dependencies were not natively annotated in the FTB. This complicates
of-course direct comparison of our work. Schluter and van Genabith [19] focus on
15We report results with the ’head’ marking strategy of Nivre and Nilsson, performing very slightly
producing treebank based LFG approximations for French, using for this a mod-ified version of part of the FTB, where they annotated functional paths for a few cases. Clergerie [7] proposes a deep symbolic parser for French, which can recover LDDs directly (including eLDDs), but currently quantitative evaluation of LDDs for this parser remains to be performed. More generally, works on LDDs do not focus specifically on eLDDs, though they are the most difficult type of LDDs. For example, Rimell et al. [17] describe a 800 English sentences corpus used for the evaluation of wide coverage parsers on unbounded dependencies, but over all eight types of unbounded dependencies only one is exclusively of eLDD type.
7
Conclusion
We have annotated cases of effectively long distance dependencies for two French treebanks, totalizing over 15000 sentences. We could verify that non local depen-dencies are very rare in actual French texts: we annotated a non local governor for 513 tokens only out of over 420000 tokens. Yet, they are massive within the occur-rences of the relative pronoun dont, and to a lesser extent que, and also frequent for the clitic en. We noted that extraction out of finite verbal clause is totally absent from the corpora we’ve annotated (one case only for over 15000 sentences), but extraction out of infinitival clause accounts for one third of the occurrences of the relative pronoun que. Further only 2 thirds of annotated eLDDs give rise to non projective links. As far as parsing is concerned, the best results, both overall and on eLDDs, are obtained when combining pseudo-projectivization and MSTParser with order 2 factors and projective decoding.
References
[1] Anne Abeillé and Nicolas Barrier. Enriching a french treebank. In Proc. of
LREC’04, Lisbon, Portugal, 2004.
[2] Aoife Cahill, Michael Burke, Ruth O’Donovan, Josef van Genabith, and Andy Way. Long-Distance Dependency Resolution in Automatically Ac-quired Wide-Coverage PCFG-Based LFG Approximations. In Proc. of
ACL’04, pages 320–327, Barcelona, Spain, 2004.
[3] Marie Candito, Benoit Crabbé, and Pascal Denis. Statistical french depen-dency parsing : Treebank conversion and first results. In Proc. of LREC’2010, Valletta, Malta, 2010.
[4] Marie Candito, Joakim Nivre, Pascal Denis, and Enrique Henestroza An-guiano. Benchmarking of statistical dependency parsers for french. In Proc.
[5] Marie Candito and Djamé Seddah. Le corpus sequoia : annotation syntaxique et exploitation pour l’adaptation d’analyseur par pont lexical. In Proc. of
TALN 2012 (in French), Grenoble, France, June 2012.
[6] Grzegorz Chrupa!a. Morfette: A tool for supervised learning of morphology. http://sites.google.com/site/morfetteweb/, 2010. Version 0.3.1. [7] Éric De La Clergerie. Convertir des dérivations TAG en dépendances . In
Proc. of TALN 2010, Montreal, Canada, 2010. ATALA.
[8] Danièle Godard. Extraction out of NP in French. Natural Language and
Linguistic Theory, 10(2):233–277, 1992.
[9] Julia Hockenmaier. Data and models for statistical parsing with Combinatory
Categorial Grammar. PhD thesis, 2003.
[10] David M. Magerman. Statistical decision-tree models for parsing. In Proc. of
ACL’95, pages 276–283, Morristown, NJ, USA, 1995.
[11] Ryan T. McDonald and Fernando C. N. Pereira. Online learning of approxi-mate dependency parsing algorithms. In Proc. of EACL’06, 2006.
[12] Yusuke Miyao and Jun’ichi Tsujii. Probabilistic disambiguation models for wide-coverage hpsg parsing. In Proc. of ACL 2005, pages 83–90, 2005. [13] Thomas Morton and Jeremy LaCivita. Wordfreak: an open tool for linguistic
annotation. In Proc. of NAACL 2003, Demos, pages 17–18, 2003.
[14] Joakim Nivre, Johan Hall, and Jens Nilsson. Maltparser: A data-driven parser-generator for dependency parsing. In Proc. of LREC’06, pages 2216– 2219, Genova, Italy, 2006.
[15] Joakim Nivre and Jens Nilsson. Pseudo-projective dependency parsing. In
Proc. of ACL’05, pages 99–106, Stroudsburg, PA, USA, 2005.
[16] Joakim Nivre, Laura Rimell, Ryan McDonald, and Carlos Gomez-Rodriguez. Evaluation of dependency parsers on unbounded dependencies. In Proc. of
COLING 2010, pages 833–841, 2010.
[17] Laura Rimell, Stephen Clark, and Mark Steedman. Unbounded dependency recovery for parser evaluation. In Proc. of EMNLP 2009, pages 813–821, 2009.
[18] Natalie Schluter and Josef van Genabith. Preparing, restructuring, and aug-menting a French Treebank: Lexicalised parsers or coherent treebanks? In
Proc. of PACLING 07, Melbourne, Australia, 2007.
[19] Natalie Schluter and Josef Van Genabith. Dependency parsing resources for french: Converting acquired lexical functional grammar f-structure annota-tions and parsing f-structures directly. 2009.
